In a move that blurs the line between storage and server vendors, Pure Storage at a Pure Accelerate 2016 conference today unveiled FlashBlade, a hyperconverged platform that makes use of RAM and Flash memory to provide both compute horsepower and all the associated storage and network services. Pure Storage also unveiled a FlashArray//m10 offering that pushes the price point for an all-Flash array to below $50,000.
Matt Kixmoeller, vice president of products for Pure Storage, says core FlashBlade is a blade server running Elasticity Scale Out Software that makes use of an object storage system to unify NV-RAM being used as a compute engine for processing data, while NAND Flash is used to provide access to anywhere from 8TB to 52TB in storage capacity. Elasticity not only makes it possible to use a single platform to process flash management, volume and storage services that are compatible with NFS or the S3 protocol exposed by the Amazon Web Services (AWS) cloud, it provides an extensible metadata engine and garbage collection system.
Holding that combined compute and storage architecture together from a transport perspective, says Kixmoeller, is a software-defined 40G Ethernet network dubbed Elastic Fabric.
Pure Storage is not only trying to leverage its base of storage customers to challenge relative newcomers such as Nutanix and SimpliVity, it will be going head to head with Hewlett-Packard Enterprise, Cisco, and the VCE unit of EMC. In fact, just about every major vendor now has at least one hyperconvergence initiative under way.
For IT organizations, the real issue is to what degree they want to embrace hyperconvergence. In some instances, offerings such as FlashBlade take a hardware centric approach. But there are also software-only approaches that don’t require an IT organization to commit to a single vendor. There are, of course, some obvious capital expense savings that can be generated by opting to standardize on a single vendor. But given the broad range of legacy systems in many IT environments, some may opt to try to drive hyperconvergence at higher levels of abstraction.
Whatever the approach, the management of compute, storage and networking is rapidly converging. That has implications for how IT infrastructure is acquired, and managed. The days when compute, storage and networking were managed in isolation are coming to a close. The only decision now is the degree to which IT organizations want that convergence to occur. In some cases, that may mean eliminating the need for IT infrastructure specialists altogether. In other cases, it may simply result in greater collaboration between specialists. Regardless of the approach, managing IT infrastructure will never be the same.